MyFlicker 2(Beta)
Based on the RiskyPlugins AI security review of the observed evidence. The review covers version 1.0.2610130, not the analysed version 1.0.2610150.
Analysis record
- Analysed
- 1 month ago 22 October 2026
- Version
- v1.0.2610150
- Artifact
- SHA256 48d7ec4a…0b815b
- Detail
- Full list
Is MyFlicker 2(Beta) safe?
MyFlicker Code Agent is an AI coding assistant for VS Code, published on OpenVSX by Kuaishou and installed about 4,558 times. It declares no permissions and no host permissions in the manifest, so its declared scope grants no workspace or network reach beyond a normal editor tab. The endpoint list attached to it includes entries such as apointapointb.zw and assetpanel.preview.toolbar.download. Both are fragments pulled out of bundled webview strings. Neither is an address the extension dials.
The most concrete network finding is NET-SOCKET_IO-extension/dist/extension/index.js-381, a socket.io client compiled into the extension entry point. If that were reaching an unknown host on activation, it would matter. An AI agent that streams suggestions from a model backend needs a persistent connection, and socket.io is the ordinary way to build one.
Most of the volume sits in dist/webview/assets, where minified copies of cytoscape.esm-DTSO7Bv0.js, mdc-DTYItulj.js and tooltip-DRcKsiUV.js each trip the same code-smell rules. Compiled graph libraries and tooltip helpers read like generic Node code. Across 5,852 matches, the malware signature category is empty and so is the secret category. No obfuscation rule fired either.
That is a scanner reading build output, which is what this extension ships by design. Nothing in the tree reads .env files, .ssh keys or VS Code secret storage, and no postinstall script runs at install time.
No findings
The analyzer reported nothing for this version.
Publisher Evidence
Lowkuaishou
Publisher identity, store signals, distribution reach, and warning signals used for context. Treat this as supporting evidence, not a clean bill of health.
12 evidence rows available.
No findings
The analyzer reported nothing for this version.
AI Security Report
AI Security Review
Evidence context: threat category none; evidence quality moderate.
MyFlicker Code Agent (budapest-vscode-extension on OpenVSX, published by Kuaishou, about 4,558 installs) is an AI coding assistant that ships a compiled extension host plus a bundled webview UI. The review produced 5,852 matches, and every one of them lands inside extension/dist/. That directory is build output: Vite and webpack chunks for the extension entry point and for the webview assets.
Filesystem and process access
The manifest declares no permissions and no host permissions. Nothing in the shipped code reads workspace paths beyond what the editor hands an extension host, and there is no child_process finding anywhere in the tree. The ten network matches are NET-SOCKET_IO and NET-FETCH hits spread across the webview chunks: extension/dist/extension/index.js:381, extension/dist/webview/assets/app-BCXWsvYK.js:1, extension/dist/webview/assets/cytoscape.esm-DTSO7Bv0.js:4 and siblings. A socket.io client in the compiled entry point is the most load-bearing piece of code here. An agent that streams completions from a hosted model needs that connection, so the capability tracks the stated purpose.
Credentials
The secret category is empty. Nothing matched a rule for .env reads, .ssh paths, .git/config, cloud credential files or VS Code's SecretStorage API. There is no postinstall script and no install-time fetch-and-execute. The 28 dependency matches are package metadata from the bundled tree.
Where the volume comes from
4,395 code-smell matches sit on minified library code. cytoscape.esm-DTSO7Bv0.js, mdc-DTYItulj.js, tooltip-DRcKsiUV.js and inline-host-Fo02oJQq.js are graph rendering, markdown and tooltip helpers; compiled output from any of them reads like generic Node code that touches fetch, process.env and crypto. The 1,419 IoC entries are the usual extractor debris: apointapointb.zw, apointcpointd.zw, 2-y.top-y.bottom-d.top, 7z.zip and asp.net are fragments of UI strings and file names scraped out of the same chunks. assetpanel.preview.toolbar.download is a dotted i18n key read as a hostname. No obfuscation rule fired.
Strongest counterargument
The socket.io client at extension/dist/extension/index.js:381 is the one finding with real reach. A persistent socket in an extension host can carry file contents out of a workspace, and the destination host is not shown in this data. What argues against that reading: the socket sits beside a webview that renders a model chat panel, the extension declares no permissions, zero secret-access patterns exist anywhere in the tree, and the endpoint list holds no plausible operator infrastructure, only string fragments such as ao.info and ar.info. A socket dialing an unrecognised domain would change the conclusion. Nothing here shows one.
Key Reasons
- All 5,852 matches fall inside extension/dist/, which is compiled webpack/Vite output rather than hand-written source.
- Zero malware signatures, zero malware family hits, zero obfuscation findings and zero secret-access matches across the entire tree.
- Manifest declares no permissions and no host permissions; no child_process or postinstall execution pattern present.
- Network findings are NET-SOCKET_IO and NET-FETCH hits in bundled libraries (cytoscape.esm-DTSO7Bv0.js, mdc-DTYItulj.js, tooltip-DRcKsiUV.js), consistent with a chat-based AI coding agent.
- Endpoint list contains string fragments such as apointapointb.zw, 7z.zip and assetpanel.preview.toolbar.download rather than real operator infrastructure.
False Positive Considerations
- 1,419 IoC entries are minified-JS substring artifacts and dotted i18n keys misread as hostnames.
- 4,395 code-smell hits fire on bundled graph, markdown and tooltip libraries in dist/webview/assets.
- Socket.io and fetch matches come from normal client code in an extension that talks to a model backend.
- 28 dependency matches are package metadata from the vendored npm tree.
Reviewed 2026-10-18 · version 1.0.2610130; recommended action: suppress false positive; model confidence 88%.
Open VSX version history
Risk trend by version
4 analyzed versions. Each point is the latest successful scan for that version; failed zero-score scans are hidden. Dates: First seen by Risky Plugins.
Pick any point on the chart, then open the Source tab to read that version's code.
Source Code Not Available
Source code is not available for this version of the extension.
About This Extension
Frequently Asked Questions
Similar Extensions
Related extensions from the same publisher or marketplace